Executive Compensation and Policy Choices at U.S. Commercial Banks
Robert DeYoung
University of Kansas, KU School of Business
1300 Sunnyside Avenue
Lawrence, KS 66045
Emma Y. Peng
Fordham University, Graduate School of Business Administration
1790 Broadway, 13th Floor
New York, NY 10019
Meng Yan
Fordham University, Graduate School of Business Administration
1790 Broadway, 13th Floor
New York, NY 10019
This draft: March 27, 2009
Presented at:
Finance Department Seminar
University of Mississippi
Abstract: In response to the huge losses taken by U.S. financial institutions in mortgage-backed
securities investments, federal legislators have proposed laws that would constrain the ability of bank
boards to freely set the size and terms of executive pay. Underlying these proposals is the belief that
corporate risk-taking can be controlled by inserting the proper incentives into executive compensation
contracts. We examine whether and how the terms of CEO compensation contracts at large, publicly
traded commercial banks between 1993 and 2006 influenced the risk-profiles of these firms, and also
whether and how the firms’ boards responded to risk-taking by changing the terms of CEO compensation
contracts. We find strong evidence linking contractual risk-taking incentives (which we proxy with
standard measures of vega and delta) to financially risky business policies, including but not limited to
increased investments in mortgage-backed securities and increased operating income from originate-to-
securitize banking activities. We also find evidence suggesting that bank boards respond in to
mismanagement and excessive risk-taking by altering the incentives in executive compensation contracts
in sensible ways. These findings suggest that (a) banking executives were to some extent aware of the
risks that they were taking, (b) these risks were taken at least partially in response to the incentives
present in their compensation contracts, but (c) government intervention to limit risk-taking incentives in
financial executive compensation contracts would at best strengthen, and at worst interfere with, the risk-
mitigation behaviors already exhibited by bank boards.
1
1. Introduction
The historic collapse of U.S. housing values, and the subsequent losses in subprime mortgages
and mortgage-backed securities, has wreaked havoc on the capital positions of financial institutions
around the world. For example, the market capitalization of the ten largest U.S. commercial banks as of
January 15, 2009 had declined by approximately $630 billion from their January 2007 levels, a collective
65% reduction in value (Reuters 2009). To encourage affected banks to continue lending—and to prevent
the most affected banks from collapsing—the U.S. Treasury injected over $300 billion in preferred and
common equity capital into commercial banking companies between Autumn 2008 and Spring 2009
through its Troubled Asset Relief Program (TARP). Large non-bank financial institutions such as
American International Group (AIG), Fannie Mae, Freddie Mac, and Bear Stearns have received even
larger amounts of aid in the form of equity injections, loans, and loss guarantees from Treasury and the
Federal Reserve.
Predictably, it did not take long for politicians and policymakers to seek to limit executive
compensation in the banking companies that received aid. To qualify for additional capital injections
from TARP in 2009, banks must limit executive compensation to $500,000 per year. More dramatically,
on March 16 President Obama instructed the Treasury Department to "pursue every legal avenue" to
block $165 million in bonuses due to executives and other financial professionals at AIG:1
"This is a corporation that finds itself in financial distress due to recklessness and
greed… Under these circumstances, it's hard to understand how derivative traders at AIG
warranted any bonuses, much less $165 million in extra pay. How do they justify this
outrage to the taxpayers who are keeping the company afloat?‖
A few days later, the House of Representatives did the President one better by passing a bill placing a
90% tax surcharge on compensation above $250,000 at any financial institution that received more than
$5 billion from TARP.
1 ―Obama Asks Geithner to Find Way to Rescind AIG Payouts,‖ Wall Street Journal Online, March 16, 2009,
http://online.wsj.com/article/SB123721970101743003.html?mod=djemalertNEWS.
2
Government interference in private executive compensation is not unprecedented in the U.S., but
it is rare.2 However, the ―bailing out‖ of large financial firms with taxpayer dollars has generated
substantial public support for intervention. The language of the President’s statement plays to this public.
His reference to ―greed‖ reflects the populist notion that executive pay in general is simply too high at
large public corporations, while ―recklessness‖ invokes an image of irresponsible (i.e., principal-agent)
behavior for which discipline should be meted out. Along these lines, many voters are asking their
representatives for punitive actions against the Wall Street executives that ―caused this mess.‖ But still,
the widest base of political support likely flows from the common belief that executive pay should reflect
the absolute level of company financial performance.
One can think of company financial performance as being driven by three key elements: the
business model in place; how well executives execute the business model; and external conditions beyond
the control of executives. These three elements converged in late-2007 for many of the largest U.S.
financial companies, with dire results. For much of the preceding two decades, large commercial banks
had been transitioning their retail business away from the traditional ―lend-and-hold‖ model which relied
on interest income generated from repeat borrower-lender relationships, and toward an ―originate-and-
sell‖ loan securitization model that relied heavily on the fee income generated by non-repeat, arms-length
financial transactions. This new business model efficiently channeled trillions of investor dollars to
mortgage borrowers, in exchange for which investors—large commercial and investment banks among
them—held mortgage-backed securities (MBS) and/or derivatives of MBS.3 This business model proved
very profitable, and generated record earnings for the commercial banking industry during most years
from the mid-1990s through the mid-2000s. But these were years of relatively benign economic
2 One of the few recent examples is the Omnibus Budget Reconciliation Act of 1993 which capped the allowable
corporate tax deduction on the salaries of a firm’s five highest paid managers to $1 million; beyond this amount,
only "qualified performance based pay" merits a deduction. The law was intended to better align executive
compensation with corporate performance. Gritsch and Snyder (2005) find that stock option compensation has
increased as a result of this change 3 These derivative securities include interest-only and principal-only instruments backed by pools of mortgages, and
more complex collateralized debt obligations which are backed by pools of MBS. Loan securitization has also
increased in credit card, auto loan, student loan, and small business credit markets; financial losses on the asset-
backed securities created in these transactions have not occurred as quickly, however, nor have they been as large, as
for MBS.
3
conditions; growing this business model absent the disciplining effects of economic stress for such a long
time period encouraged excesses.4 The collapse of the housing bubble exposed these excesses, most
notable among them the subprime MBS with investment-grade ratings that performed so abysmally upon
the collapse of the housing bubble in 2007.
In retrospect, it is clear that managers at commercial and investment banks committed a number
of fundamental risk-management mistakes. Portfolios were over-weighted in MBS, as bank managers
(like some famous economists) underestimated the covariances of regional housing price movements.5
The financial leverage used against these investments was often excessive, both on the balance sheet and
in off-balance sheet investment vehicles. And these levered portfolios of long-term assets were often
financed with short-term debt, suggesting that managers forgot or simply ignored the key lessons of the
1980s savings and loan crisis.
Arguments in favor of government-imposed limits on executive compensation at these
institutions must ultimately be judged by two questions: Did the incentives embedded in executive
compensation contracts at U.S. financial firms during the 1990s and 2000s cause or contribute to this
record of financial mismanagement? And if so, is government regulation of executive compensation
necessary to curtail future episodes of mismanagement, or do these firms (i.e., their compensation
committees) naturally respond to mismanagement and excessive risk-taking by adjusting the performance
incentives in executive contracts? In this paper, we devise and implement empirical tests that address
each of these questions. We find strong (albeit preliminary) statistical evidence linking contractual risk-
taking incentives at commercial banks to financially risky business policies—including but not limited to
risks associated with the originate-to-sell business model—and we find suggestive evidence that bank
4 Between 1991 and 2007 there was only a single, relatively shallow recession. And ironically, the consumer
spending generally credited for the mildness of the 2001 recession was made possible by mortgage securitization,
which permitted homeowners to more readily access the equity that would previously been locked up in their homes. 5 This view was not limited to investors in home real estate securities. Alan Greenspan famously stated during
congressional testimony that upward pressure on home prices was largely a regional phenomenon and that
nationwide declines in home prices were unlikely. (Testimony to Congress on July 20, 2005.)
4
boards respond to mismanagement and excessive risk-taking by altering the incentives in executive
compensation contracts in sensible ways.
We use market proxies—so-called delta and vega—to capture the incentives present in executive
compensation contracts between 1993 and 2006 at large publicly traded U.S. commercial banking
companies. Pay-performance sensitivity, or delta, measures the semi-elasticity of CEO compensation to
changes in the firm’s stock price. Pay-risk sensitivity, or vega, measures the change in CEO
compensation with respect to changes in stock price volatility. Over the past two decades total CEO
compensation at the largest U.S. commercial banks has been similar to that paid to CEOs at the largest
U.S. industrial companies (see Figure 1). However, the incentive structures embedded in the contracts of
bank CEOs permanently diverged from those of non-bank CEOs around 2000, when bank CEO pay
became substantially more sensitive to stock price volatility (see Figure 2). Perhaps coincidently, this
increase in CEO return to risk-taking occurred at approximately the same time that annual increases in
U.S. housing prices began to accelerate (see Figure 4). We test whether and to what degree vega and
delta are determinants of banks’ policy decisions regarding the composition of bank income, loan mix,
credit quality, securities investments, deposit mix, financial leverage, off-balance sheet activities, and
asset growth rates. Following Coles, et al. (2006), we use a three-stage least squares (3SLS) estimation
approach to control for the potential endogeneity of executive compensation contracts to bank policy
choices. This not only improves the accuracy of our estimates, but also allows us to test whether and how
compensation committees respond to managers’ policy decisions by adjusting the delta and vega in
compensation packages.
We find plentiful evidence that bank business policies are influenced by the incentives present in
CEO compensation contracts. On average, banks in which CEOs have been given stronger incentives to
take risk (high-vega banks) generate a larger percentage of their incomes from noninterest activities—
including but not limited to securitization activities (e.g., loan origination, loan securitization, and loan
servicing fees)—and invest a larger percentage of their assets in private (i.e., subprime or otherwise non-
conforming) mortgage securitizations. High-vega banks also take more credit risk, use more financial
5
leverage, and rely more on non-core deposit funding. Importantly, the data indicate that this bundle of
relatively risky business policies exposes high-vega banks primarily to systematic risk, and thus
exacerbates the financial stress often experienced by banks during macroeconomic downturns. In
contrast, banks in which CEOs are less incented to take risk (low-vega banks) tend to invest larger
portions of their assets in on-balance sheet loan portfolios and relatively safer government-sponsored
pass-through mortgage securitizations. Holding vega constant, tying CEO compensation more closely to
stock price (high-delta contracts) is nearly always associated with policy choices in the opposite direction
of those made at high-vega banks. This suggests that, at least at commercial banking companies, high-
delta compensation packages can induce risk-averse policy incentives that temper or balance the risk-
inducing effects of high-vega compensation packages (or vice versa).
Our results also suggest, at least for some business policies, that bank boards are monitoring and
responding to risk outcomes with sensible adjustments to compensation contracts. Pay-risk sensitivity
(vega) tends to be set lower at banks exposed to the most credit risk, e.g., those invested in large amounts
of nonperforming loans or risky (non-agency) mortgage-backed securities. Consistent with this, pay-
performance sensitivity (delta) tends to be set higher at banks invested in risky MBS as well as at banks
that exhibit high amounts of systematic or idiosyncratic risk; this suggests attempts to induce greater CEO
risk aversion. However, for other business policies boards appear to be reinforcing risky policy choices—
presumably because boards perceived these policies as reasonable rather than excessive risks, or because
boards did not perceive these policies to be risky during our sample period (e.g., mortgage-backed
securities?). Pay-risk sensitivity tends to be set higher at banks that are generating large amounts of
income from non-traditional activities (e.g., brokerage, insurance, investment banking), financially
levering their off-balance sheet activities, or exhibiting high amounts of systematic risk.
The remainder of the paper is organized as follows. Section 2 discusses the transactions banking
model central to the expansion of mortgage credit in the U.S. and how the adoption of this model has
affected bank financial performance. Section 3 reviews the relevant literature on executive compensation
and risk-taking. Section 4 presents our empirical model and test methodology and states the broad
6
hypotheses linking executive compensation and risk-taking. Section 5 describes our data, defines the
variables we use to specify the model, and presents variable-specific hypotheses. Section 6 reports our
(preliminary) empirical results. Section 7 summarizes and concludes.
2. Transactions banking
The top commercial banking companies in the U.S. have grown immensely larger over the past
two decades. During the mid-1980s only Citibank had more than $100 billion in assets; by the mid-2000s
nearly 20 U.S. banking companies had more than $100 billion, and three exceeded $1 trillion. And even
these burgeoning asset figures understate the expansion of the largest U.S. banks, because the industry
simultaneously experienced a doubling of noninterest income from activities largely unrelated to bank
assets.
What permitted this staggering increase in bank size? Deregulation allowed banks to expand
their geographic footprints and offer non-banking products such as investment banking, brokerage, and
insurance sales and underwriting.6 But why was deregulation necessary? Innovations in financial
markets and information technologies set in motion a process of disintermediation that threatened to make
the heavily regulated U.S. banking sector obsolete. The advent of money market mutual funds, 401K
plans, discount brokerage, and other new options for depositors and savers was depriving banks of their
most important sources of funds, while the expansion of commercial paper, high-yield debt, and OTC
stock markets was providing banks’ core business customers with ready substitutes for bank loans.
The banks most successful at coping with disintermediation created an entirely new retail banking
business model that took advantage of looser regulation, new channels of information, and deeper
financial markets. ―Transactions banking‖ embraces financial disintermediation. Banks use their
expertise in loan underwriting to originate loans, but instead of issuing deposits to fund these loans on-
6 In 1994 the U.S. Congress passed the Riegle-Neal Interstate Banking and Branching Efficiency Act, which
effectively repealed the McFadden Act at the national level and harmonized the patchwork of state-by-state banking
and branching rules. In 1999 Congress passed the Graham-Leach-Bliley Financial Services Modernization Act,
which effectively repealed the Glass-Steagall Act by granting broad-based securities and insurance powers to
commercial banking companies.
7
balance sheet, they (or their investment bank partners) issue asset-based securities to fund these loans in
off-balance sheet loan securitization pools. Loan securitizations are investment trusts that purchase
existing home mortgage loans (or auto loans, or credit card receivables) from banks, using funds raised by
selling ―mortgage-backed securities‖ (MBSs) to third-party investors. This process allows banks to sell
their otherwise illiquid loans to the securitization, and use the proceeds of these sales to fund additional
loans—in a sense, recycling bank capital. Banks earn noninterest income from loan origination fees, loan
securitization fees, and loan servicing fees, while the loan interest and principal repayments are shared by
the MBS investors. The banking system, however, is not shed of the credit risk associated with the
securitized loans. Depending on the riskiness of the loans in the pool, the originating or securitizing
banks may hold a portion of the MBSs themselves or provide recourse agreements to MBS investors.
And the investors in MBS can be banks and other financial institutions that want exposure to the risks and
returns of diversified pools of mortgages (or other retail loans) without having to generate these loans
themselves.
The growth in securitized mortgage lending was facilitated in large part by government-
sponsored enterprises (GSEs) such as Ginnie Mae, Fannie Mae, Freddie Mac. Well over half of the
residential mortgage debt in the U.S. is securitized by, held in the portfolios of, or guaranteed by these
three institutions. Most of the MBSs issued by these GSEs are relatively safe and easy-to-understand
―pass-through‖ securities: the pooled mortgages are either backed by government guarantees, private
insurance, or large down payments, and the interest and principal cash flows are shared equally by the
investors.7 But investors in other types of MBS can bear substantial amounts of risk. Private (non-GSE)
mortgage-backed securities are backed by pools of non-conforming loans that carry additional risk for a
7 Because Fannie Mae and Freddie Mac initially securitized or held only conforming mortgages (non-jumbo first
mortgages with either 20% down payments or private mortgage insurance), they were permitted to operate with very
little capital; moreover, their lines of credit at the U.S. Treasury created the perception that they were ―too-big-to-
fail,‖ which gave them a funding advantage over private-sector mortgage securitizers. But in response to political
pressure during the early 2000s, both Fannie and Freddie began purchasing subprime MBSs. As these investments
soured and the GSEs reached the verge of insolvency, the Treasury Department made good on its ―implicit
government guarantee‖ by injecting equity funding and nationalizing ownership of the two GSEs.
8
wide range of reasons: they are large (jumbo) loans, they have low down payments, borrows provided
incomplete documentation of income (low-doc loans), or borrowers had poor credit ratings (subprime
loans). And structured mortgage-backed securities—derivative MBS products that are backed by pools of
other MBS—decompose the underlying mortgage interest and principal repayments into ―tranches‖
according to financial needs and risk appetites of investors: interest-only tranches, principal-only
tranches, and tranches that are either more or less exposed to prepayment risk or credit risk.
The transactions banking model has yielded large production and financing efficiencies for banks
that use it, and by sharing credit risk with investors outside of the banking system has increased access to
credit for millions of households and small businesses. But mismanagement of this lending technology
was a key contributor to the bubble in U.S. housing markets during the 2000s and the subsequent turmoil
in world financial markets.
Transactions banks gain access to enormous economies of scale (Hughes et al. 1996, Rossi 1998).
These scale economies are associated with the collection and analysis of the ―hard,‖ quantifiable borrower
information central to the automated lending processes used to evaluate, originate, and pool large volumes
of retail loans (Stein 2002). But transactions banks all have access to the same information (e.g., credit
scores) and all produce non-differentiated financial commodities such as mortgage loans and credit card
loans, so price competition is intense and profit margins are tight. Hence, transactions banks have strong
incentives to grow larger in order to exploit further unit cost reductions. Acquiring other banks has been
the most efficient way to achieve increased scale; over 10,000 bank charters have been merged out of
existence since 1980. Once transactions banks have exhausted their external growth options, internal
growth requires increasing the number of loan originations, which creates incentives to relax lending
standards and make loans to less creditworthy borrowers. This incentive to write increasingly risky loans
is amplified by the fact that, in its purest form, the transactions banking model separates loan
underwriting from the monitoring and bearing of credit risk. Although pooling loans does to some extent
reduce risk via diversification, MBS investors bear the bulk of the credit risk. And given the information
problems associated with pools comprised of hundreds or thousands of individual loans, as well as the
9
complexity of the expected cash flows from some structured MBS products, investors simply cede the
task of evaluating credit risk to third-party securities rating firms.
Earnings at large banking companies have become more reliant on noninterest income over
time—not just from their transactions banking activities, but also from other nontraditional lines of
business made accessible by deregulation such as securities underwriting and brokerage, and from selling
backup lines of credit that enable their business clients to issue their own debt securities. DeYoung and
Roland (2001) argue that increased reliance on noninterest has altered the risk-return profiles of banks.
For example, compare (a) the fee income a bank receives by originating and securitizing a mortgage loan
to (b) the interest income a bank receives by originating a small business loan and holding it in its loan
portfolio. The former is a non-repeat transaction with the borrower, and the volume of this business is
sensitive to housing market volatility and mortgage interest rates—that is, systematic risk. The latter is a
long-term relationship that both sides have an interest in preserving, which will continue to generate
interest income (and perhaps fee income as well) into the future. In a similar fashion, fees charged in
securities brokerage are typically based on the value of assets sold or under management; again,
predominantly systematic risk. Moreover, activities that generate noninterest income increase banks’
operating leverage—the production functions for most of these activities are dominated by fixed costs,
e.g., personnel expenses—as well as their effective financial leverage if these activities do not leave
footprint on the balance sheet. Leverage amplifies revenue volatility into even greater earnings volatility.
A number of empirical studies have investigated the volatility of noninterest income at banks and
its effect on risk. DeYoung and Roland (2001) show that (non-deposit-related) fee income is associated
with higher revenue volatility, higher operating leverage, and higher earnings volatility at U.S.
commercial banks. DeYoung and Rice (2004b) find that marginal increases in non-interest income are
associated with a worsening of banks’ risk-return trade-off. Stiroh (2004a, 2004b) finds no evidence of
diversification gains at banks that combine interest and non-interest income. Choi, et al. (2006) find that
noninterest income at commercial banking companies in 42 different countries is strongly and positively
related to systematic risk. Clark, et al. (2007) emphasize how the increasingly retail-focused strategies of
10
large U.S. banking companies expose these banks to economic and business cycle volatility. Elysiani and
Wang (2008) demonstrate that noninterest income makes it more difficult for analysts to forecast the
quarterly earnings of banking companies.
The sub-prime mortgage crisis provides an illustration of the income volatility associated with
fee-driven transactions banking. While the headlines in the financial press have justifiably dwelled on the
over $2 trillion of capital losses suffered by banks and other investors in sub-prime mortgage-backed
securities, transactions banking companies that originated, serviced, and securitized mortgages have
experienced material, and in some cases crippling, reductions in fee income as investor demand for new
MBS dried up and household demand for both new and existing houses declined. Total industry
noninterest income fell from 43% to 38% of operating income between 2006 and the first three quarters
of 2008, the largest two-year decline since the mid-1970s. Many of the largest financial institutions with
non-diversified, ―mono-line‖ mortgage banking strategies failed (e.g., American Home Mortgage, New
Century Financial, Countrywide Financial, Washington Mutual) due to the combined impact of
plummeting fee income and large losses in their portfolios of subprime mortgages and mortgage-backed
securities.
3. Executive Compensation
Why does executive compensation affect firms’ policy choices? According to agency theory,
CEO compensation shapes managerial incentives, and delta and vega are two important measures of such
managerial incentives (Core and Guay 2002). Delta, or pay-performance sensitivity, measures the change
in the dollar value of CEO wealth for a 1% change in stock price. Vega, or pay-risk sensitivity, captures
the change in the dollar value of CEO wealth for a 0.01 change in stock return volatility. The impact of
high delta is two-fold. On the one hand, high-delta contracts tie managerial wealth to shareholder value
by paying managers with shares of firm stock, in attempts to reduce conflicts of interest between
managers and shareholders (Jensen and Meckling 1976; Morck et al. 1988; McConnell and Servaes 1990;
Berger et al. 1997). On the other hand, high-delta contracts concentrate managerial wealth in the shares
11
of the firm, providing incentives for poorly diversified, risk-averse managers to pass up positive NPV
projects that increase firm risk. High-vega contracts seek to mitigate managerial risk-aversion by paying
managers in stock options, which should make risk more valuable to managers by increasing their pay-
risk sensitivity (Jensen and Meckling 1976; Smith and Stulz 1985). Studies of industrial firms provide
evidence that high-vega contracts encourage riskier policy choices while high-delta contracts encourage
less risky policy choices (Knopf et al. 2002; Rogers 2002; Nam et al. 2003; Coles et al. 2006).
Executive compensation in the banking industry has not traditionally been structured to
encourage risk-taking (Smith and Watts 1992; Houston and James 1995). However, as discussed in
section 2, industry deregulation expanded banks’ investment opportunities, allowing them to expand into
new geographic markets and provide non-banking financial services such as investment banking,
securities brokerage, and insurance sales and underwriting. To motivate their CEOs to take advantage of
these growth options, bank boards dramatically reshaped executive contracts. Although total executive
compensation increased no faster than in other industries (see Figure 1), the composition of bank CEO
pay became more equity-based. Total equity compensation (including both options grants and restricted
stock grants) at large banking companies increased from 29% of total compensation in 1992 to more than
50% in 2000, before slipping back to about 40% in 2005.8 More importantly, the changing composition
of bank CEO compensation resulted in larger deltas and vegas (see Figures 2 and 3), which indicate that
bank CEO wealth has become more sensitive to stock price and stock return volatility.
The terms of executive compensation contracts may not be exogenous; in response to managers’
policy choices—and the risk profiles implied by those choices—boards may take actions to alter
managerial incentives. Guay (1999) suggests that firms with more R&D expenditures are more likely to
have high-vega contracts. Coles et al. (2006) further show that vega increases in R&D expenditures and
leverage, while vega decreases in investments in plant assets and firm focus. Similar evidence has been
8 Full data is available upon request. The non-monotonicity can be attributed mainly to option grants, whose values
and percentages peaked in 2000 and declined steadily since then. One explanation for these declines is the
lackluster stock performance after 2000 compared to the stock market boom period of 1994-2000. Another
explanation is the beginning of option expensing around 2002 (Carter et al. 2007).
12
compiled for banking companies (Crawford et al. 1995; Hubbard and Palia 1995). Chen et al. (2006)
document that, in the post-deregulation era, option-based compensation at banks is positively related to
market-based measures of bank risk.
The main focus in this study is on vega and the interplay of pay-risk sensitivity and bank policy
choices. However, we include delta as an important control variable in our tests, which allows us to also
observe the relationships between delta and bank policy choices. There are mixed theories on delta’s
impact on risk-taking. Smith and Stulz (1985) show that high delta increases the risk aversion of
undiversified managers, causing managers to reduce firm risk. But John and John (1993) argue that high-
delta contracts can provide incentives to shift risk to debt-holders, causing managers to take excessive risk
on behalf of shareholders (i.e., themselves). This is a legitimate concern here, since asset substitution
problems can be more serious in banks where the majority of debt is typically in the form of deposit
contracts guaranteed by the Federal Deposit Insurance Corporation (FDIC). Since one bank may have a
very different mix of vega and delta than another bank, it is essential to control for the various underlying
effects of delta when we examine the relationships between vega and banks’ policy choices.
4. Methodology
Following the literature, we expect that (a) bank CEOs with stronger risk-taking incentives will
implement riskier business policies and (b) bank boards will adjust compensation packages in response to
CEO policy choices and the outcomes of those policies. Hence, compensation contracts and business
policy choices are simultaneously determined, and ordinary least squares (OLS) regressions of policy
measures on contemporaneous compensation incentives may generate biased parameter estimates. To
address the endogeneity issue, we use two different estimation methods in our tests: (1) OLS with lagged
compensation variables, and (2) three-stage least squares (3SLS) estimation.9
9 Although 3SLS generates more efficient coefficient estimates in large samples, the small-sample properties of
3SLS are not well understood. Because of this, we also used two-stage least squares (2SLS) in our analyses (not
shown here, available upon request). The untabulated 2SLS results do not change our inferences.
13
4.1. Ordinary Least Squares
To account for the likelihood that policy choices will have feedback effects on the terms of
compensation contracts, we estimate the following OLS business policy model (firm subscripts
suppressed for convenience):
Policyt = f(ln(Vegat-1), ln(Deltat-1), ln(TA t-1), ln(MBt-1), Controls t-1), (1)
where ln(Vega t-1) and ln(Deltat-1) are the natural logs of one-period lagged vega and delta, and ln(TA t-1)
and Ln(MBt-1) are the natural logs of total assets and the market-to-book ratio of equity measured at the
beginning of the year. We predict that ln(Vegat-1) is positively related to policies that increase bank risk.
However, we have no unambiguous expectation for the impact of ln(Deltat-1), as there is no firm
consensus in prior studies on whether high-delta contracts encourage managerial risk-aversion or
managerial risk-taking behavior on net. We expect that larger banks and banks with better investment
opportunities will be more likely to engage in nontraditional banking activities that generate more volatile
income streams and higher market risk exposures. Controlst-1 includes unique control variables in each of
the specific business policy regressions; these business policy-specific control variables are listed in
Appendix B.
4.2. Three-Stage Least Squares
To address the joint determination of business policies and compensation incentives, we need to
isolate the impact of business policy choices on CEO compensation structure from the impact of CEO
compensation structure on business policy choices. Therefore, we estimate simultaneous equation
systems using 3SLS, which combines two-stage least squares (2SLS) and seemingly unrelated regression
(SUR) techniques. Specifically, we estimate the following system of equations:
Policyt = f(ln(Vegat), ln(Deltat), ln(TA t-1), ln(MBt-1), Controls t-1) (2)
14
ln(Vegat) = f(Policyt, ln(TA t-1), ln(MBt-1), ln(Salaryt) (3)
ln(Deltat) = f(Policyt, ln(Vegat), ln(TA t-1), ln(MBt-1), Cash_Balancet-1, Tenuret) (4)
where ln(Salaryt) is the natural log of CEO salary for the year, Cash_Balancet-1 is the percentage of assets
held in cash at the beginning of the year, and Tenuret is the number of years in the CEO’s term. The
control variables in the bank policy regressions (2) are the same as those in the OLS regressions. We
include ln(Salaryt) in the vega regression (3) as a proxy for managerial risk-aversion (Core and Guay,
1999; Guay, 1999; Coles et al., 2006). We expect that bank CEOs with larger amounts of fixed pay will
be more willing to take on risk. We expect a negative coefficient on Cash_Balances and a positive
coefficient on Tenure in the delta regression (4). Prior studies have shown that compensation committees
at cash-constrained firms are more likely to provide CEOs with high-delta contracts; high-delta contracts
are also more likely as managerial ability becomes less uncertain or when the approach of CEO retirement
creates horizon problems.
5. Data
Our sample is based on the intersection of the ExecuComp database and the Federal Reserve Y-
9C Bank Holding Company database from 1993 through 2006. ExecuComp reports top executive
compensation information extracted from annual proxy statements; we estimate our key variables delta
and vega from these data. The FR Y-9Cs contain quarterly financial data that cumulates over the calendar
year; we extract business policy variables, as well as most of our control variables, from the year-end
December 31 reports. We start with 141 banks (1,057 bank-years) in the ExecuComp database.10
The
process of lagging variables, merging the two databases, and estimating delta and vega reduces this to 139
10
Our banking sample includes ExecuComp observations with the SIC code of 6020.
15
banks (974 bank-years). As shown in Table 1, the number of banks in each year ranges from a low of 61
in 1993 to a high of 75 in 1998. The sample includes a total of 216 CEOs.11
Table 2 presents descriptive statistics on CEO characteristics, business policy measures, and
additional bank characteristics. Following Core and Guay (2002), we use the ―one-year approximation‖
method to generate annual estimates of vega and delta.12
Vega has a mean (median) of $149,351
($50,917), and delta has a mean (median) of $616,410 ($280,742). In other words, the average bank CEO
enjoys an increase of $149,351 in his/her equity portfolio for a 0.01 increase in stock return volatility, and
an increase of $616,410 for a 1% increase in stock price. Since the two variables have large standard
deviations and are skewed to the right, we use log transformations to produce more symmetric data
distributions. At the average bank in our sample, the CEO has 9 years of tenure in the position, and earns
$4.55 million in total annual compensation—approximately $710,000 in salary, $970,000 in bonus, $1.62
million in option grants, and $610,000 in restricted stock.
We examine twenty-eight policy measures that are expected to affect bank risk. Descriptive
statistics for these variables appear in Panel B of Table 2. We include four measures of noninterest
income. Nonint is total noninterest income, Nonint_Less is total noninterest income less fiduciary income
and deposit service charges, Nonint_Sec is noninterest income from loan securitization and servicing, and
Nonint_Nontrad is noninterest income from nontraditional banking activities such as investment banking,
brokerage, trading, venture capital, and insurance underwriting. Each is scaled by net operating income
(i.e., noninterest income plus interest income minus interest expense). Based on the extant banking
literature (e.g., DeYoung and Roland 2001), we expect each of the noninterest measures to be risk-
increasing and thus positively related to vega. (Because the general compensation literature provides
11
As discussed in the following section, we do not used firm fixed effects in our estimations. Because our results
are based mainly on cross-sectional variation in the data, changes in bank CEOs do not pose a critical problem. 12
We value CEO stock options using the Black-Scholes (1973) model modified by Merton (1973) to account for
dividends payouts. Vega is the partial derivative of the option value with respect to stock-return volatility, multiplied
by 0.01 times the number of options. Delta equals delta from options plus delta from stock holdings. Delta from
options is the partial derivative of the option value with respect to stock price, multiplied by 1% of the current stock
price times the number of options. Delta from stock holdings is simply the product of 1% of the current stock price
and the number of shares.
16
mixed results on the relationship between delta and risk-taking, we will not establish any expectations in
this section for the direction of the delta-policy associations.)
We include six loan portfolio measures. Loans is total loans and leases held in portfolio,
Commer_Loans is commercial and industrial loans, and Commer_Real_Loans is commercial real estate
loans. Each is scaled by total assets. Because large banks have increasingly shifted credit risk off of their
balance sheets and on to their income statements (Stiroh 2004, 2006), we expect investments in the loan
portfolio—holding credit risk constant—to be negatively related to vega. We also measure loan quality.
Alloc_loans is allocation for loan and lease losses, Prov_loans is provision for loan and lease losses, and
Charge_loans is net charge-offs and recoveries on loan and lease losses. Each is scaled by total assets.
Since low quality loans expose banks to greater credit risk, we expect these three (ex post) measures of
credit risk to be positively related to vega.
We include five bank financing measures. Noncore_Deposit is the percentage of bank assets
funded by noncore deposits, such as time deposits of $100,000 or more. Short_Funds is the percentage of
bank assets funded by purchased fed funds and repurchase agreements. Funding long-term investments
with short-term and/or unstable liabilities increases both interest rate risk and liquidity risk at banks.
Therefore, we expect these two variables to be positively related to vega. We include three measures of
financial leverage as well: Equity_Mult is the simple equity multiplier, total assets divided by book equity
capital; Equity_Mult_Off equals the sum of total assets and off-balance sheet asset equivalents (e.g.,
unused commitments and letters of credit) divided by book equity capital; and Equity_Mult_Flows equals
net operating income divided by book equity capital. We expect financial leverage to be positively
related to vega.
We include two measures to capture banks’ expansion and growth. Growth is the natural log of
the annual increase in total assets, and Intangible is the percentage of bank assets that are intangible.
Both of these measures are positively related to asset acquisitions. Because banks that grow more
aggressively will tend to exhibit greater operational risks, we expect these two variables to be positively
related to vega. We also include the notional value (Deriv_Notional) and fair value (Deriv_FV) of
17
derivatives contracts held for trading purposes, both of which are scaled by total assets. Conceptually, we
would expect derivatives trading to be positively related to vega. However, the data to which we have
access does not indicate the net long versus short exposures in these contracts. Furthermore, the vast
majority of derivatives contracts in the banking industry are held at just a small handful of banks (Minton
et al. 2006). For both of these reasons, our expectations regarding the derivatives-vega relationship are
not strong ones.
We include six measures of investment in mortgage backed securities (MBS): the amortized cost
and fair value of private securitizations (MBS_Private_BV and MBS_Private_FV), the amortized cost and
fair value of structured mortgage-backed products (MBS_ Struct _BV and MBS_ Struct _FV), and the
amortized cost and fair value of agency pass-through securities (MBS_ Pass_BV and MBS_Private_FV).
Each is scaled by total assets. As discussed above, pass-through MBS are backed by insured and/or
conforming mortgages, so we expect these investments to be negatively related to vega. In contrast, we
expect investments in complex structured MBS products and low credit-quality private MBS products to
be positively related to vega.
Finally, we include three measures of market-based risk. Risk is the standard deviation of daily
stock returns, Beta is the slope coefficient estimated from a standard one-factor market model, and
Idiosyn_Risk is the standard deviation of the market model residuals. Prior studies suggest that higher
vega leads to higher total risk (Coles et al. 2006), and that the value of executive stock options increases
with systematic risk after controlling for total risk (Meulbroek, 2001; Duan and Wei 2005). As result, we
expect both Risk and Beta to be positively related to vega. We have a weaker expectation that
Idiosyn_Risk will also be positively related to vega.
Panel C of Table 2 reports descriptive statistics for the control variables in each of our models.
As discussed above, some of these variables appear as controls in all versions of the model—asset size
and the market-to-book ratio in the business policy regressions; CEO salary in the vega regressions; and
cash balances and CEO tenure in the delta regressions—while others are used to help identify selected
versions of the business policy regressions.
18
Table 3 reports Pearson and Spearman correlations among the 28 policy variables; the three risk
measures Risk, Beta, and Idiosyn_Risk; and the compensation incentives Vega and Delta. Consistent with
our expectations, fee income, commercial loans, problem loans, noncore deposits, financial leverage,
asset expansion, derivative contracts, and private MBS are all positively correlated with the risk
measures. Also as expected, noninterest income, problem loans, financial leverage, asset expansion,
derivative contracts, and private MBS all show positive correlations with ln(Vega) and ln(Delta), while
on-balance sheet lending and structured MBS are negatively related to ln(Vega) and ln(Delta). Consistent
with prior studies, ln(Vega) and ln(Delta) are positively related to Beta. Importantly, noninterest income,
problem loans, financial leverage, intangible assets, and derivative contracts are all positively correlated
to Beta—although these bivariate correlations do not control for the influence of other factors, they are
suggestive that transactions banks (noninterest income, financial leverage) may be especially vulnerable
to economic downturns.
6. Results
Our 1993-2006 data are an annual panel—however, because the chief variables of interest
(business policy, vega, delta) exhibit little variation over time at the firm level, fixed effects estimation is
not appropriate.13
Instead, we estimate both model (1) and model (2,3,4) using two alternative
techniques. First, we apply ordinary least squares (OLS) with year fixed effects to the pooled time series-
cross section data panel. The year dummies absorb annual variations in business policy and/or
compensation practices that are common to all firms. Still, given that the variation in our data is primarily
cross-sectional, simply pooling the data in this fashion will over-represent the power of our statistical
tests. Our solution is to use annual OLS cross-section estimation with Fama-McBeth estimators, a
13
When we applied a firm fixed effects approach to the data (not shown here) we found few statistically
relationships between the policy variables and the compensation variables. The many strong and economically
sensible correlations between these variables in Table 3, taken together with the disappearance of these relationships
in firm fixed effects regressions, leads us to conclude that most of the variation in our data is cross-sectional.
19
flexible approach that permits all coefficients to vary each year. We report the results of both estimation
procedures in the tables below.
We present selected results from the OLS estimations of model (1) in Table 4. (We do not
discuss the coefficients on the control variables Z in this preliminary version of the paper. The interested
reader can consult the full regression results reported in Appendix C.) Although the results are not always
robust and a number of our hypothesized compensation-policy relationships do not obtain, the statistically
strongest and most robust results tell an economically sensible story. Banks with high pay-risk sensitivity
(high vega) tend to have higher than average amounts of systematic risk, and their policy choices reflect
this. High-vega banks rely more on noninterest income—which as we argue above exposes banks to
systematic risk (DeYoung and Roland 2001, Choi, et al. 2006, Clark et al. 2007)—and hold greater
amounts of intangible assets. Because these business policies make the bank more opaque, they are
difficult for analysts to value (Elysiani and Wang 2008) and hence will tend to increase stock price
volatility for these banks. In contrast, banks with high pay-performance sensitivity (high delta) exhibit
higher idiosyncratic risk, and their policy choice reflect this. High-delta banks invest a high percentage of
their assets in loans, and as a result their returns are more likely to follow the firms and sectors to which
they lend rather than the market average. The high reliance on short-term funding and the high
percentage of problem loans at these banks are both consistent with policies aimed at expanding the loan
portfolio. High-delta banks also tend to grow their assets faster; in a stock market that values earnings
growth, this policy is consistent with high pay-performance sensitivity. Finally, there is a sensible
symmetry in these results: high-vega (high-delta) CEOs choose policies that make their banks more (less)
reliant on noninterest income from securitization and non-traditional activities, and less (more) reliant on
traditional portfolio lending.
Of course, model (1) does not account for the potential endogeneity of executive compensation
contracts to the policy choices made by managers, and as such the coefficient estimates reported in Table
4 may be biased. We present selected results from the model (2,3,4) estimations in Table 5. Again, we
20
estimate this model two ways: pooled 3SLS estimation with year dummies, and annual 3SLS cross-
section estimation with Fama-McBeth estimators.
Identifying the three-equation 3SLS model is a challenge, because each version of the system is
based on a different business policy variable. The vega equation (3) and the delta equation (4) are
specified the same way in all 28 of the regression systems, but the right-hand side specification of the
business policy equation (2) must vary by definition. In a few of the systems, we use a business policy
variable from one of the other systems as an identifying instrument in the business policy regression (2).
All of the system estimations reported in Table 5 are identified based on the following straightforward
test: each equation in the system contains at least one statistically significant coefficient on a variable that
has been excluded from the other two equations in the system. A list of these identifying right-hand side
variables appears in the Appendix B.
The results from the 3SLS business policy regressions tell the same basic story as do the single-
equation OLS results above, but controlling for endogeneity generates much more robust results and
brings the story into sharper focus. High-vega banks have high amounts of systematic risk and, consistent
with this exposure, generate substantially larger portions of their operating income from noninterest
income. For example, using the Fama-McBeth estimator, a one standard deviation increase in vega
increases the noninterest share of total operating income by about 38%. (The statistically significant and
positive impact of delta on securitization-based noninterest income is smaller—however, we observe this
variable in only six of the 14 years in our data, which makes estimation and inference more difficult.)
High-vega banks make a number of unambiguously risk-increasing policy choices. Loan quality
is lower (greater provisions for loan losses, more net loan charge-offs), suggesting high amounts of credit
risk. A greater portion of assets are funded with non-core deposits, which increases both interest rate risk
and liquidity risk. A smaller portion of activities are backed by equity finance (equity multiplier_flows
and equity multiplier_off), which increases insolvency risk. And high-vega banks are relatively more
likely to invest in high-risk mortgage-backed securities (MBS_private) than in low-risk agency
securitizations (MBS_pass-through). The two results that seeming contradict this pattern are the
21
significant negative associations of vega and holdings of commercial real estate loans and derivatives
securities. However, these two results should be interpreted with caution. Commercial real estate lending
is a highly competitive but relatively safe business except during real estate downturns when loan defaults
can increase dramatically; but because real estate downturns had previously been regional phenomena,
large banks can run well-diversified portfolios of these loans. (This is in contrast to small banks that hold
non-diversified portfolios of local commercial real estate loans which, as a result, historically comprise
the largest percentage of bank insolvencies.) As to derivatives, since the regulatory data only allow us to
measure aggregate derivatives holdings, we cannot measure the proportions of net long-versus-short
exposures that could expose banks to losses (or gains) from interest rate or currency movements.
Consistent with the incentives they face, high-delta banks (holding vega constant) make policy
choices that expand bank activities in areas that are well understood by investors. For example, they loan
out a substantially larger than average percentage of their assets; using the Fama-McBeth estimator, a one
standard deviation increase in delta increases the lending share of total assets by 15%. Within the loan
portfolio, high-delta banks substitute commercial real estate loans for general (non-real estate) business
loans, while outside the loan portfolio they generate a smaller than average portion of their income from
difficult-to-value non-interest activities. These results suggest that high pay-performance sensitivity
creates incentives for bank executives to run relatively traditional banking models. The one result that is
inconsistent with this picture is the positive association between delta and private mortgage
securitizations.
The other two equations in the system capture the ―feedback‖ from managers’ policy choices to
the compensation incentives that the board provides management. The results displayed in the last four
columns of Table 5 suggest that compensation committees adjust both delta and vega in a coordinated
fashion, in attempts curb management from taking excessive risk. For instance, in the vega equations, the
loan quality, private MBS, and structured MBS policy variables all carry significantly negative
coefficients; this is strong evidence that bank boards attempt to moderate risk taking and reduce policy
excesses by weakening CEO risk-taking incentives. Consistent with this, delta tends to be set higher for
22
banks with large amounts of non-core deposit funding, banks that invested in private and structured MBS,
with systematic risk, and with idiosyncratic risk; this suggests that bank boards attempt to influence CEOs
to make more risk-averse policy decisions by increasing their pay-performance sensitivity.
For other policy choices, however, bank boards appear to be reinforcing risky policy choices. For
example, boards tend to increase vega when the share of noninterest income in total income is large and
when banks finance these cash flows with high amounts of financial leverage. Why would potentially
risk-increasing policy choices be greeted with higher risk-taking incentives? This is a sensible response if
we remember the conventional wisdom in the banking community during most of our sample period that
fee-based activities were risk-reducing at banks (DeYoung and Roland 2001); given these incorrect
beliefs, rational boards would seek to increase risk-taking incentives at noninterest-intensive banks, even
to the point of further levering these activities. Another potential explanation is that boards did
understand that noninterest income tends to increase the bank’s systematic risk, and increased vega to
incent management to increase the volatility of stock price (i.e., reduce the correlation between bank
returns and market returns). This interpretation would also explain the significant positive coefficient on
systematic risk in the vega equation.
Finally, the significantly negative coefficients on the three loan variables in the delta regression
are not clearly intuitive, and we offer the following speculative interpretation: Because loans (and
especially business loans) are the most traditional of commercial bank business policies, the market
understands these policies and will give them high risk-adjusted values; given the positive market
reactions to these policies, bank managers with high deltas could easily over-invest in loans, causing the
board to reduce pay-performance sensitivity.
One must be careful when interpreting the compensation ―feedback‖ results in the vega and delta
equations. The coefficient estimates on the policy variables are largely capturing cross-sectional variation
in the data—the coefficients do not represent inter-temporal reactions within given firms. For example,
the negative coefficients on loan charge-offs in the vega equation indicate that high loan losses relative to
the industry will tend to elicit less risk-performance sensitive contracts. The coefficient estimates on delta
23
in the vega equations (on vega in the delta equation) are interpreted similarly. Note that when these
coefficients are statistically significant, they always carry a positive sign—that is, bank boards tend to
alter the terms of CEO compensation in order to jointly increase or decrease delta and vega. This is
consistent with the use of pay-risk sensitivity to offset the unwanted managerial risk-aversion side effects
of pay-performance sensitivity.
7. Summary, Discussion, and Conclusions
The huge losses suffered by large U.S. financial institutions that created and invested in risky
mortgage-backed securities, and the nearly equally huge government subsidies spent to keep these firms
functioning and afloat, have raised the public’s ire. True to form, politicians have responded with a
variety of schemes to limit the pay of the financial executives who ―got us into this mess.‖ While much
of this is political theater—in time, a substantial portion of the government loans and capital injections are
likely to be paid back, and ex post facto sanctions on employee pay and bonuses are unlikely to withstand
constitutional scrutiny—the episode has raised public consciousness and increased the likelihood of a
more permanent role for government in monitoring and determining executive pay in publicly traded
companies. In 2005, Representative Barney Frank introduced ―The Protection Against Executive
Compensation Abuse Act,‖ which called for increased public disclosure of executive pay—including the
targets for short-term and long-term performance incentives—and would have required shareholder
approval of executive compensation contracts. The bill did not progress beyond initial stages. However,
Frank and others are considering new legislation that would link executive pay to company performance,
and such a bill would have a better chance for passage in today’s environment.14
Underlying the efforts to control executive pay is the belief that corporate risk-taking can be
controlled by inserting the proper incentives into executive compensation contracts. We examine whether
and how the terms of CEO compensation contracts at large, publicly traded commercial banks between
1993 and 2006 influenced the risk-profiles of these firms, and also whether and how the firms’ boards
14
―Cuomo, Frank Seek to Link Executive Pay, Performance,‖ Susanne Craig, Wall Street Journal, March 13, 2009.
24
responded to risk-taking by changing the terms of CEO compensation contracts. We find strong evidence
linking contractual risk-taking incentives (which we proxy with standard measures of vega and delta) to
financially risky business policies, including but not limited to increased investments in mortgage-backed
securities and increased operating income from originate-to-securitize banking activities. We also find
evidence suggesting that bank boards respond in to mismanagement and excessive risk-taking by altering
the incentives in executive compensation contracts in sensible ways.
We draw three broad conclusions from these findings. First, banking executives were to some
extent aware of the risks that they were taking by investing in risky MBS and by increasing the proportion
of their banks’ revenues generated by originate-and-securitize banking activities. Our findings run
contrary to the claim that banks were misled by over-optimistic ratings on MBS (although such claims
may be valid for less sophisticated investors). Second, banking executives endorsed these risky business
policies at least partially in response to the incentives present in their compensation contracts. This lends
legitimacy to arguments for government intervention to limit contractual risk-taking incentives for
executives at systemically important financial institutions. Third, government intervention to limit risk-
taking incentives in financial executive compensation contracts would at best strengthen, and at worst
interfere with, the risk-mitigation behaviors already being exhibited by bank boards. For those prone to
policy activism, our findings that bank boards respond to risky policy choices by sometimes attempting to
constrain risk-taking, while in other instances attempting to encourage additional risk-taking, could be
seen as a justification for intervention. But the terms of optimal contract incentives are likely to vary
from firm to firm and from CEO to CEO, while government prescriptions almost by necessity tend to be
one size fits all. Furthermore, the contractual incentives that we test here were designed by boards to
mitigate principal-agent problems on the behalf of shareholders, while contractual incentives imposed via
regulation are presumably aimed at providing public goods (i.e., financial market stability, fairness) and
could work far differently.
In interpreting our 1993-2006 findings, it is important to realize that bank managers were making
their policy decisions conditional not only on the incentives structured into their compensation
25
agreements, but also conditional on their beliefs regarding the risk-return tradeoffs associated with their
various policy options. The 2007-2009 financial crisis is likely to have changed managers’ understanding
of risks and returns in some lines of business. For example, the housing downturn revealed that many
mortgage-backed securities were far riskier than suggested by either their third-party ratings or their
contractual yields; managers’ beliefs about the risk-return tradeoffs inherent in MBS are likely to have
changed post-downturn. Thus, our 2006 tests reflect bank managers’ pre-crisis beliefs about the risk-
return qualities of MBS, and may only imperfectly capture how their business policy choices will react to
contractual risk-taking incentives post-crisis. Similarly, our estimates are based on the incomplete pre-
crisis understanding of these risk-return tradeoffs by bank boards and compensation committees. Thus,
one must be careful when drawing inferences about optimal post-crisis policy based on our pre-crisis
results. Proposals to foster macroeconomic stability by rolling back banking powers may be misguided,
because informed post-crisis managers will arguably be better able to implement those powers effectively.
And proposals to constrain risk-taking by constraining the ability of bank boards to set the terms of
executive compensation may also be misguided, because informed post-crises boards will arguably be
better able to determine efficient incentives.
Finally, we note that none of these conclusions are meant to extend to non-banking firms.
Commercial banks are subject to supervisory monitoring that, if not explicitly, implicitly creates extra
pressure for boards to mitigate risk-taking managerial behaviors. Moreover, the level and types of risk
taken by bank executives, and endorsed by bank compensation committees, during the 1990s and 2000s,
to some large extent are special to the newness of the transactions banking business model and the
incomplete understanding of the risks inherent in that model and the products it created.
26
References
Berger, P.G., Ofek, E., Yermack, D.L., 1997. Managerial entrenchment and capital structure decisions.
Journal of Finance 52, 1411-1438.
Carter, M.E., Lynch, L.J., Tuna, I., 2007. The role of accounting in the design of CEO equity
compensation. The Accounting Review 82, 327-357.
Chen, C.R., Steiner, T.L., Whyte, A.M., 2006. Does stock option-based executive compensation induce
risk-taking? An analysis of the banking industry. Journal of Banking and Finance 30, 915-945.
Choi, S., DeYoung, R., Hasan, I., 2006. ―Risk, Return, and Noninterest Income at Commercial Banks:
Cross-Country Evidence, unpublished working paper, Rensselaer Polytechic Institute.
Clark, T., Dick, A., Hirtle, B., Stiroh, K., Williams R., 2007. The role of retail banking in the U.S.
banking industry: risk, return, and industry structure. Economic Policy Review – Federal Reserve
Bank of New York 13, 39-56.
Coles, J.L., Daniel, N.D., Naveen, L., 2006. Managerial incentives and risk-taking. Journal of Financial
Economics 79, 431-468.
Core, J., Guay, W., 1999. The use of equity grants to manage optimal equity incentive levels. Journal of
Accounting and Economics 28, 151–184.
Core, J., Guay, W., 2002. Estimating the value of employee stock option portfolios and their sensitivities
to price and volatility. Journal of Accounting Research 40, 613-630.
Crawford, A.J., Ezzell, J.R., Miles, J.A., 1995. Bank CEO pay-performance relations and the effects of
deregulation. Journal of Business 68, 231-256.
DeYoung, R., Roland, K.P., 2001. Product mix and earnings volatility at commercial banks: evidence
from a degree of total leverage model. Journal of Financial Intermediation 10, 54-84.
DeYoung, R., Rice, T., 2004b. Noninterest income and financial performance at U.S. commercial banks.
The Financial Review 39, 101-127.
Duan, J., Wei J., 2005. Executive stock options and incentive effects due to systematic risk. Journal of
Banking and Finance 29, 1185-1211.
Elysiani, E., Wang, Y., 2008. Non-Interest Income Diversification and Information Asymmetry of Bank
Holding Companies, unpublished working paper, Temple University.
Gritsch, M., Snyder, T.C., 2005. The impact of current tax policy on CEO stock option compensation: a
quantile analysis. Academy of Accounting and Financial Studies Journal, September.
Guay, W., 1999. The sensitivity of CEO wealth to equity risk: an analysis of the magnitude and
determinants. Journal of Financial Economics 53, 43-71.
Houston, J.F., James, C., 1995. CEO compensation and bank risk: is compensation in banking structured
to promote risk taking. Journal of Monetary Economics 36, 405-431.
27
Hubbard, R.G., Palia, D., 1995. Executive pay and performance: evidence from the U.S. banking
industry. Journal of Financial Economics 39, 105-130.
Hughes, J. P., Lang, W., Mester, L. J., Moon, C., 1996. Efficient banking under interstate branching.
Journal of Money, Credit, and Banking 28, 1045-1071.
Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: managerial behavior, agency costs, and
ownership structure. Journal of Financial Economics 3, 305-360.
John, T.A., John, K., 1993. Top-management compensation and capital structure. Journal of Finance 48,
949-974.
Knopf, J.D., Nam, J., Thornton, J.H. Jr., 2002. The volatility and price sensitivities of managerial stock
option portfolios and corporate hedging. Journal of Finance 57, 801-814.
McConnell, J.J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. Journal
of Financial Economics 27, 595-612.
Meulbroek, L., 2001. The efficiency of equity-linked compensation: understanding the full cost of
awarding executive stock options. Financial Management 30, 5-44.
Minton, B., Stulz R.M., Williamson R., 2006. How Much do Banks use Credit Derivatives to Reduce
Risk? Fisher College of Business Working Paper No. 2006-03-001, June 2006.
Morck, R., Shleifer, A., Vishny, R., 1988. Management ownership and market valuation: an empirical
analysis. Journal of Financial Economics 20, 293-315.
Nam, J., Ottoo, R.E., Thornton Jr., J.H. 2003. The effect of managerial incentives to bear risk on
corporate investment and R&D investment. The Financial Review 38, 77-101.
Rogers, D.A., 2002. Does executive portfolio structure affect risk management? CEO risk-taking
incentives and corporate derivative usage. Journal of Banking and Finance 26, 271-295.
Rossi, C. V., 1998. Mortgage banking cost structure: resolving an enigma. Journal of Economics and
Business 50, 219-234.
Smith, C.W., Stulz, R.M., 1985. The determinants of firms’ hedging policies. Journal of Financial and
Quantitative Analysis 20, 391-405.
Smith, C.W., Watts, R.L., 1992. The investment opportunity set and corporate financing, dividend, and
compensation policies. Journal of Financial Economics 32, 263-293.
Stein, J. C., 2002. Information production and capital allocation: decentralized versus hierarchical firms.
Journal of Finance 57, 1891-1921.
Stiroh, K., 2004a. Do community banks benefit from diversification? Journal of Financial Services
Research 25, 135-160.
Stiroh, K., 2004b. Diversification in banking: is noninterest income the answer? Journal of Money,
Credit, and Banking 36, 853-882.
28
Stiroh, K., 2006. New evidence on the determinants of bank risk. Journal of Financial Service Research
30, 237-263.
29
Table 1
Year Membership of Sample Banks (N=974, 1993-2006)
Year No. of
Observations
1993 61
1994 73
1995 66
1996 71
1997 71
1998 75
1999 73
2000 72
2001 63
2002 71
2003 74
2004 67
2005 71
2006 66
Total 974
30
Table 2
Descriptive Statistics (1993-2006)
Panel A: CEO Characteristics
# of
Obs Mean
Standard
Deviation
25th
Percentile Median
75th
Percentile
Vega ($000s) 974 149.351 248.861 19.603 50.917 158.083
ln(Vega) 974 3.948 1.623 3.025 3.950 5.069
Delta ($000s) 974 616.410 905.716 112.823 280.742 710.537
ln(Delta) 974 5.637 1.324 4.735 5.641 6.567
Salary ($000s) 974 710.023 285.042 515.703 687.308 887.160
ln(Salary) 974 6.489 0.401 6.247 6.534 6.789
Bonus ($000s) 974 971.519 1414.090 207.160 476.197 1107.710
Option_Grants ($000s) 971 1621.510 2612.350 150.044 635.783 1968.510
Rstock_Grants ($000s) 908 606.218 1677.260 0.000 0.000 385.608
Total_Compensation ($000s) 971 4549.960 5304.250 1412.900 2461.070 5301.590
Tenure 973 9.026 6.674 4 7 13
Panel B: Business Policy Measures
# of
Obs
# of
Years Mean
Standard
Deviation
25th
Percentile Median
75th
Percentile
Nonint 974 14 0.361 0.162 0.248 0.322 0.430
Nonint_Less 974 14 0.206 0.139 0.113 0.166 0.252
Nonint_Sec 412 6 0.015 0.034 0.000 0.002 0.014
Nonint_Nontrad 412 6 0.066 0.078 0.015 0.041 0.082
Loans 974 14 0.619 0.145 0.578 0.657 0.706
Commer_Loans 974 14 0.145 0.076 0.094 0.137 0.184
Commer_Real_Loans 974 14 0.145 0.100 0.076 0.124 0.198
Alloc_loans 974 14 0.010 0.004 0.008 0.009 0.012
Prov_loans 974 14 0.003 0.003 0.001 0.002 0.004
Charge_loans 974 14 0.003 0.003 0.001 0.002 0.004
Noncore_Deposit 974 14 0.094 0.068 0.050 0.073 0.115
Short_Funds 974 14 0.074 0.059 0.035 0.060 0.101
Equity_Mult 974 14 12.363 2.546 10.642 12.206 13.648
Equity_Mult_Flows 974 14 0.688 0.203 0.556 0.659 0.772
Equity_Mult_Off 974 14 17.333 6.528 13.672 16.263 19.275
Growth 974 14 0.128 0.153 0.038 0.092 0.177
Intangible 974 14 0.019 0.017 0.006 0.013 0.024
Deriv_Notional 839 11 1.181 4.430 0.000 0.001 0.144
Deriv_FV 839 11 0.032 0.123 0.000 0.000 0.003
MBS_Private_BV 912 13 0.016 0.034 0.000 0.001 0.017
MBS_Private_FV 912 13 0.016 0.034 0.000 0.001 0.016
MBS_Struct_BV 912 13 0.039 0.044 0.004 0.026 0.059
MBS_Struct_FV 912 13 0.039 0.043 0.004 0.025 0.058
MBS_Pass_BV 912 13 0.068 0.065 0.023 0.048 0.092
MBS_Pass_FV 912 13 0.068 0.065 0.024 0.047 0.092
Risk 974 14 0.017 0.006 0.013 0.016 0.020
Beta 974 14 0.921 0.346 0.683 0.894 1.150
Idiosyn_Risk 974 14 0.015 0.006 0.011 0.013 0.017
31
Table 2 (Continued)
Panel C: Bank Characteristics
N Mean
Standard
Deviation
25th
Percentile Median
75th
Percentile
Net_Oper_Rev ($000s) 974 3,019,320 5,712,160 342,417 958,449 2,827,081
Total_assets ($000s) 974 54,266,555 109,286,316 6,642,681 17,533,980 48,624,000
ln(TA) 974 16.690 1.345 15.594 16.529 17.609
MB 973 2.375 0.975 1.673 2.156 2.811
ln(MB) 973 1.179 0.264 0.983 1.149 1.338
Cash_Balance 954 0.056 0.045 0.031 0.044 0.065
Deposit 929 0.704 0.106 0.640 0.708 0.782
Core_deposit 927 0.612 0.115 0.550 0.618 0.694
Equity_Ratio 974 0.083 0.016 0.072 0.081 0.092
Net_Interest_Margin 927 0.046 0.011 0.041 0.046 0.052
Foreign_Deposit 929 0.041 0.081 0.000 0.005 0.040
One_Yr_Gap 974 0.200 0.154 0.096 0.208 0.313
Nonint_Less/Beg(TA) 974 0.015 0.018 0.006 0.010 0.016
Trading_Rev/Beg(TA) 774 0.001 0.002 0.000 0.000 0.001
All variable definitions are in Appendix A.
32
Table 3
Correlation (N=974, 1993-2006)
Panel A: Pearson Coefficients
Risk Beta Idiosyn_Risk Ln(Vega) Ln(Delta)
Nonint 0.089*** 0.260*** -0.006 0.406*** 0.353***
Nonint_Less 0.117*** 0.265*** 0.031 0.374*** 0.355***
Nonint_Sec 0.044 0.078 -0.037 0.320*** 0.230***
Nonint_Nontrad -0.047 0.097*** -0.133*** 0.324*** 0.186***
Loans -0.153*** -0.169*** -0.118*** -0.099*** -0.093***
Commer_Loans 0.080*** 0.025 0.085*** -0.008 -0.040
Commer_Real_Loans -0.099*** -0.107*** -0.077** -0.155*** -0.181***
Alloc_loans 0.011 0.072*** 0.035 -0.046 -0.029
Prov_loans 0.301*** 0.148*** 0.240*** 0.172*** 0.161***
Charge_loans 0.181*** 0.273*** 0.117*** 0.251*** 0.226***
Noncore_Deposit 0.155*** -0.028 0.169*** -0.003*** 0.023
Short_Funds 0.026 0.031 0.015 0.027*** 0.123
Equity_Mult 0.164*** 0.001 0.175*** -0.042 0.009
Equity_Mult_Flows 0.213*** 0.136*** 0.187*** 0.108*** 0.191***
Equity_Mult_Off 0.232*** 0.303*** 0.167*** 0.219*** 0.318***
Growth 0.101*** 0.003 0.114*** -0.068** 0.045
Intangible -0.123*** 0.184*** -0.207*** 0.320*** 0.287***
Deriv_Notional 0.059 0.226*** -0.019 0.286*** 0.239***
Deriv_FV 0.069** 0.228*** -0.012 0.275*** 0.218***
MBS_Private_BV 0.070** 0.040 0.069** 0.078*** 0.145***
MBS_Private_FV 0.071** 0.040 0.069** 0.080*** 0.146***
MBS_Struct_BV 0.024 -0.058* 0.035 -0.112*** -0.125***
MBS_Struct_FV 0.029 -0.058* 0.038 -0.110*** -0.122***
MBS_Pass_BV 0.042 -0.027 0.048 -0.011 -0.017
MBS_Pass_FV 0.045 -0.027 0.048 -0.009 -0.014
Risk 1 0.156*** 0.940*** 0.058* 0.060*
Beta 0.156*** 1 -0.043 0.244*** 0.205***
Idiosyn_Risk 0.940*** -0.043 1 -0.075** -0.047
33
Table 3 (Continued)
Panel B: Spearman Coefficients
Risk Beta Idiosyn_Risk Ln(Vega) Ln(Delta)
Nonint 0.009 0.261*** -0.075** 0.452*** 0.369***
Nonint_Less 0.018 0.241*** -0.070** 0.418*** 0.354***
Nonint_Sec -0.115** -0.017 -0.146*** 0.283*** 0.200***
Nonint_Nontrad -0.202*** -0.011 -0.269*** 0.436*** 0.277***
Loans -0.136*** -0.119*** -0.143*** -0.039 -0.047
Commer_Loans 0.066** 0.018 0.074** 0.002 -0.008
Commer_Real_Loans -0.101*** -0.222*** -0.102*** -0.197*** -0.220***
Alloc_loans 0.037 0.042 0.072** -0.032 -0.040
Prov_loans 0.237*** 0.050 0.182*** 0.160*** 0.116***
Charge_loans 0.107*** 0.187*** 0.050 0.279*** 0.197***
Noncore_Deposit 0.124*** -0.150*** 0.118*** -0.035 -0.018
Short_Funds -0.056* -0.051 -0.046 -0.017 0.097***
Equity_Mult 0.150*** 0.001 0.180*** -0.074** 0.009
Equity_Mult_Flows 0.193*** 0.128*** 0.191*** 0.152*** 0.181***
Equity_Mult_Off 0.140*** 0.230*** 0.116*** 0.215*** 0.245***
Growth 0.069** -0.011 0.100*** -0.106*** 0.057*
Intangible -0.128*** 0.184*** -0.255*** 0.332*** 0.282***
Deriv_Notional -0.093*** 0.254*** -0.187*** 0.483*** 0.426***
Deriv_FV -0.089*** 0.256*** -0.192*** 0.494*** 0.423***
MBS_Private_BV -0.030 0.007 -0.058* 0.117*** 0.173***
MBS_Private_FV -0.028 0.008 -0.057* 0.119*** 0.176***
MBS_Struct_BV -0.043 -0.107*** -0.029 -0.144*** -0.160***
MBS_Struct_FV -0.041 -0.108*** -0.028 -0.143*** -0.158***
MBS_Pass_BV 0.017 -0.047 0.012 0.008 -0.026
MBS_Pass_FV 0.020 -0.047 0.013 0.010 -0.024
Risk 1 0.135*** 0.927*** 0.051 0.043
Beta 0.135*** 1 -0.061* 0.273*** 0.223***
Idiosyn_Risk 0.927*** -0.061* 1 -0.110*** -0.086***
All variable definitions are in Appendix A.
34
Table 4
Selected OLS Results
OLS Robust Error OLS Fama McBeth
Obs ln(Vega) t-1 ln(Delta) t-1 Years ln(Vega) t-1 ln(Delta) t-1
Nonint
773 0.023*** -0.011** 14 0.0217*** -0.006
[0.004] [0.005] [0.002] [0.006]
Nonint_Less 773 0.011*** 0.009 14 0.013*** 0.013
[0.004] [0.006] [0.003] [0.008]
Nonint_Sec 357 0.003*** 0.0001 6 0.003* 0.001
[0.001] [0.002] [0.001] [0.002]
Nonint_Nontrad 357 0.004** -0.012*** 6 0.004 -0.012***
[0.002] [0.003] [0.002] [0.002]
Loans 773 -0.010** 0.009* 14 -0.007* 0.014***
[0.004] [0.006] [0.004] [0.005]
Commer_Loans 773 0.00002 -0.005* 14 0.000 -0.003
[0.002] [0.003] [0.001] [0.002]
Commer_Real_Loans 773 -0.003 -0.002 14 -0.004 0.003
[0.003] [0.004] [0.003] [0.003]
Alloc_Loans 799 0.000 0.001*** 14 0.000** 0.001***
[0.000] [0.000] [0.000] [0.000]
Prov_Loans 799 0.000 0.000** 14 0.000 0.000**
[0.000] [0.000] [0.000] [0.000]
Charge_Loans 799 0.000 0.000** 14 0.000*** 0.001**
[0.000] [0.000] [0.000] [0.000]
Noncore_deposit 799 0.002 0.005** 14 0.003* 0.003
[0.002] [0.002] [0.002] [0.002]
Short_Funds 799 -0.006*** 0.009*** 14 -0.005*** 0.006**
[0.002] [0.002] [0.002] [0.002]
Equity_Mult 799 -0.163*** 0.110 14 -0.124** -0.010
[0.061] [0.098] [0.042] [0.111]
Equity_Mult_flows 799 0.003 -0.008 14 0.004 -0.007
[0.005] [0.006] [0.006] [0.005]
Equity_Mult_off 799 -0.185 0.596*** 14 0.014 0.387
[0.155] [0.197] [0.133] [0.222]
Growth 771 -0.003 0.012** 14 -0.005 0.011**
[0.005] [0.005] [0.005] [0.004]
Intangible 773 0.001*** 0.001 14 0.001*** 0.001
[0.000] [0.001] [0.000] [0.001]
Deriv_Notional 646 -0.072 0.131 11 -0.068 0.152**
[0.080] [0.102] [0.051] [0.061]
Deriv_FV 646 -0.001 0.002 11 0.000 0.003*
[0.002] [0.003] [0.002] [0.001]
Year Dummies Yes
35
Table 4 (Continued)
OLS Robust Error OLS Fama McBeth
Obs ln(Vega) t-1 ln(Delta) t-1 Years ln(Vega) t-1 ln(Delta) t-1
MBS_Private_BV 742 0.000 0.004*** 13 -0.001 0.004**
[0.001] [0.001] [0.001] [0.001]
MBS_Private_FV 742 0.000 0.004*** 13 -0.001 0.004**
[0.001] [0.001] [0.001] [0.001]
MBS_Struct_BV 742 -0.001 -0.002 13 -0.002** -0.002*
[0.001] [0.002] [0.001] [0.001]
MBS_Struct_FV 742 -0.001 -0.002 13 -0.002** -0.002*
[0.001] [0.002] [0.001] [0.001]
MBS_Pass_BV 742 0.001 0.002 13 -0.000 -0.001
[0.002] [0.002] [0.002] [0.002]
MBS_Pass_FV 742 0.001 0.002 13 -0.000 -0.001
[0.002] [0.002] [0.002] [0.002]
Risk 799 0.000 0.000 14 0.000*** 0.000
[0.000] [0.000] [0.000] [0.000]
Beta 799 0.01 -0.019 14 0.033*** -0.009
[0.009] [0.012] [0.007] [0.015]
Idiosyn_Risk 799 0.000 0.000** 14 0.000 0.000**
[0.000] [0.000] [0.000] [0.000]
Year Dummies Yes
All variable definitions are in Appendix A.
36
Table 5
3SLS Results (1993-2006)
Equation Business policy Vega Delta
Right-hand Side Variable ln(Vega) ln(Delta) Business policy ln(Delta) Business Policy ln(Vega)
Policy Variable Year Obs Estimation Technique
Nonint
926 3SLS Robust Error
0.102*** -0.098*** 1.241*** 0.064 1.077** 0.610***
[0.018] [0.016] [0.475] [0.124] [0.496] [0.070]
14
3SLS Fama-McBeth 0.104*** -0.077*** 1.579*** 0.155 1.461 0.556***
[0.022] [0.019] [0.544] [0.145] [0.853] [0.117]
Nonint_Less
926 3SLS Robust Error
0.039*** -0.033*** 0.938 -0.014 0.850 0.609***
[0.014] [0.012] [0.788] [0.124] [0.669] [0.070]
14
3SLS Fama-McBeth 0.024 0.008 1.333* 0.047 2.211* 0.434**
[0.032] [0.024] [0.696] [0.132] [1.162] [0.147]
Nonint_Sec
410 3SLS Robust Error
0.020** -0.013** 6.655 0.216 8.309 0.593***
[0.008] [0.005] [6.821] [0.159] [7.341] [0.157]
6
3SLS Fama-McBeth 0.015* -0.009* 5.624 0.214 17.731 0.416
[0.006] [0.003] [7.055] [0.191] [13.158] [0.389]
Nonint_Nontrad
410 3SLS Robust Error
0.031** -0.034*** 6.302*** 0.423*** -3.656 0.831***
[0.015] [0.010] [1.712] [0.141] [1.832] [0.138]
6
3SLS Fama-McBeth -0.006 -0.003 5.162*** 0.327* 0.535 0.446
[0.026] [0.019] [0.927] [0.148] [2.366] [0.290]
37
Table 5 (Continued)
Equation Business policy Vega Delta
Right-hand Side Variable ln(Vega) ln(Delta) Business policy ln(Delta) Business Policy ln(Vega)
Policy Variable Year Obs Estimation Technique
Loans
924 3SLS Robust Error
-0.038** 0.041*** -1.089** 0.043 -1.783*** 0.761***
[0.016] [0.014] [0.498] [0.116] [0.612] [0.080]
14
3SLS Fama-McBeth -0.058 0.060*** -1.286 0.186* -3.573* 0.601***
[0.036] [0.015] [0.889] [0.104] [1.756] [0.128]
Commer_Loans
924 3SLS Robust Error
0.021*** -0.026*** 1.268 0.054 -6.380*** 0.713***
[0.009] [0.007] [2.237] [0.145] [1.634] [0.080]
14
3SLS Fama-McBeth 0.009 -0.025* 0.486 0.235 -4.068* 0.661***
[0.010] [0.012] [1.894] [0.151] [2.107] [0.076]
Commer_Real_Loans
924 3SLS Robust Error
-0.057*** 0.051*** -4.626*** 0.224** -4.987*** 0.667***
[0.012] [0.010] [1.299] [0.116] [2.133] [0.080]
14
3SLS Fama-McBeth -0.055*** 0.047*** -2.329 0.242** -6.652** 0.473***
[0.014] [0.008] [1.632] [0.099] [2.406] [0.139]
Alloc_Loans 952
3SLS Robust Error 0.000 0.000 -46.144** -0.123 -17.388 0.680***
[0.000] [0.000] [23.218] [0.136] [18.883] [0.075]
14 3SLS Fama-McBeth -0.0001 -0.0001 -52.360*** -0.078 -4.513 0.644***
[0.0003] [0.0002] [13.022] [0.131] [27.269] [0.115]
Prov_Loans 952 3SLS Robust Error
0.002*** -0.001** -119.420* -0.157 -38.148 0.682***
[0.000] [0.000] [71.910] [0.154] [56.932] [0.133]
14 3SLS Fama-McBeth 0.001** -0.0002 -209.968 -0.066 305.463 0.736***
[0.0003] [0.0003] [121.28] [0.177] [281.672] [0.217]
Charge_Loans
952 3SLS Robust Error
0.002*** -0.001*** -138.885* -0.266 -32.747 0.721***
[0.000] [0.000] [75.664] [0.182] [50.168] [0.101]
14
3SLS Fama-McBeth 0.001*** -0.0002 -117.907* -0.094 282.02 0.818***
[0.0004] [0.0003] [61.121] [0.541] [229.229] [0.238]
38
Table 5 (Continued)
Equation Business policy Vega Delta
Right-hand Side Variable ln(Vega) ln(Delta) Business policy ln(Delta) Business Policy ln(Vega)
Policy Variable Year Obs Estimation Technique
Noncore_Deposit
952 3SLS Robust Error
0.017** -0.006 6.666** 0.040 9.174*** 0.509***
[0.006] [0.006] [2.825] [0.110] [3.316] [0.115]
14
3SLS Fama-McBeth 0.015** -0.009 4.384 -0.065 13.889*** 0.302
[0.007] [0.009] [3.668] [0.155] [4.381] [0.180]
Short_Funds
952 3SLS Robust Error
0.008 -0.001 -6.514* 0.040 -11.964*** 0.721***
[0.007] [0.006] [4.006] [0.116] [4.684] [0.096]
14
3SLS Fama-McBeth -0.005 -0.0005 -3.694 0.238 31.550 0.067
[0.012] [0.012] [5.301] [0.147] [29.410] [0.371]
Equity_Mult
952 3SLS Robust Error
-0.232 -0.141 -0.124* 0.022 -0.281*** 0.532***
[0.273] [0.231] [0.073] [0.113] [0.112] [0.128]
14
3SLS Fama-McBeth 0.225 -0.931** -0.023 0.298* -0.163 0.713*
[0.348] [0.423] [0.130] [0.147] [0.542] [0.394]
Equity_Mult_Flows 952 3SLS Robust Error
0.058*** -0.058*** 1.710*** 0.056 1.147*** 0.627***
[0.019] [0.016] [0.420] [0.114] [0.386] [0.081]
14 3SLS Fama-McBeth 0.053** -0.057 2.210*** 0.019 1.686** 0.426***
[0.021] [0.034] [0.543] [0.134] [0.762] [0.118]
Equity_Mult_Off 952 3SLS Robust Error
3.102*** -1.350*** 0.047*** 0.049 0.033*** 0.583***
[0.640] [0.551] [0.012] [0.116] [0.010] [0.086]
14 3SLS Fama-McBeth 2.811*** -1.425** 0.013 -0.010 0.063 0.470***
[0.668] [0.558] [0.026] [0.133] [0.036] [0.155]
Growth 922 3SLS Robust Error
0.015 0.015 -11.280*** 0.150 2.448 0.554***
[0.017] [0.014] [3.131] [0.132] [4.790] [0.110]
14 3SLS Fama-McBeth 0.004 0.007 -3.057 0.180 2.838 0.522***
[0.020] [0.021] [3.421] [0.140] [2.928] [0.098]
Intangible
924 3SLS Robust Error
0.005*** -0.003* -12.097 -0.004 -16.786 0.742***
[0.002] [0.001] [18.509] [0.136] [15.322] [0.096]
14
3SLS Fama-McBeth 0.002 -0.0004 -15.017 0.083 -24.498 0.776***
[0.002] [0.001] [17.297] [0.132] [20.719] [0.153]
39
Table 5 (Continued)
Equation Business policy Vega Delta
Right-hand Side Variable ln(Vega) ln(Delta) Business policy ln(Delta) Business Policy ln(Vega)
Policy Variable Year Obs Estimation Technique
Deriv_Notional
646 3SLS Robust Error
-2.956*** 0.851* -0.007 0.052 0.020 0.663***
[0.575] [0.520] [0.023] [0.162] [0.017] [0.092]
11
3SLS Fama-McBeth -2.583*** 0.993 0.021 0.234 -0.004 0.588***
[0.645] [0.562] [0.020] [0.133] [0.014] [0.134]
Deriv_FV
646 3SLS Robust Error
-0.082*** 0.022 -0.469 0.058 0.798 0.668***
[0.016] [0.015] [0.824] [0.162] [0.595] [0.092]
11
3SLS Fama-McBeth -0.063*** 0.021 1.053 0.214 -0.110 0.588***
[0.016] [0.014] [0.745] [0.127] [0.518] [0.134]
MBS_Private_BV
870 3SLS Robust Error
0.003 0.015*** -26.256*** 0.354* 17.663*** 0.442***
[0.004] [0.004] [9.420] [0.197] [5.088] [0.093]
13
3SLS Fama-McBeth 0.007** 0.009*** -15.052 -0.175 30.311*** 0.043
[0.003] [0.003] [16.848] [0.258] [8.281] [0.186]
MBS_Private_FV 870 3SLS Robust Error
0.003 0.015*** -26.423*** 0.356* 17.766*** 0.442***
[0.004] [0.004] [9.464] [0.198] [5.109] [0.093]
13 3SLS Fama-McBeth 0.007** 0.009*** -15.199 -0.175 30.269*** 0.044
[0.003] [0.003] [16.833] [0.256] [8.312] [0.185]
MBS_Struct_BV 870 3SLS Robust Error
-0.003 0.001 -20.801*** 0.000 17.998*** 0.704***
0.005] [0.004] [4.875] [0.138] [4.865] [0.091]
13 3SLS Fama-McBeth -0.002 -0.0001 -7.516 0.020 10.223** 0.542***
[0.004] [0.003] [4.422] [0.121] [4.384] [0.116]
MBS_Struct_FV 870 3SLS Robust Error
-0.003 0.001 -20.767*** -0.001 18.001*** 0.703***
[0.005] [0.004] [4.896] [0.138] [4.862] [0.091]
13 3SLS Fama-McBeth -0.002 -0.0001 -7.731 0.019 10.256** 0.543***
[0.004] [0.003] [4.454] [0.121] [4.424] [0.117]
MBS_Pass_BV 870 3SLS Robust Error
-0.022*** 0.024*** -0.156 0.001 2.577*** 0.639***
[0.008] [0.007] [1.565] [0.136] [1.089] [0.069]
13 3SLS Fama-McBeth -0.026** 0.014 -6.841* 0.086 -3.603 0.365
[0.010] [0.010] [3.723] [0.162] [4.464] [0.207]
MBS_Pass_FV
870 3SLS Robust Error
-0.023*** 0.024*** -0.225 0.004 2.603** 0.641***
[0.008] [0.007] [1.562] [0.136] [1.086] [0.069]
13
3SLS Fama-McBeth -0.027** 0.015 -6.741* 0.096 -3.461 0.363
[0.009] [0.010] [3.723] [0.161] [4.458] [0.207]
40
Table 5 (Continued)
Equation Business policy Vega Delta
Right-hand Side Variable ln(Vega) ln(Delta) Business policy ln(Delta) Business Policy ln(Vega)
Policy Variable Year Obs Estimation Technique
Risk
952 3SLS Robust Error
0.001*** 0.000 95.283*** 0.005 87.872*** 0.581***
[0.000] [0.000] [32.197] [0.117] [28.729] [0.084]
14
3SLS Fama-McBeth 0.00005 0.0006 30.813 0.124 70.710* 0.483***
[0.0005] [0.0005] [59.114] [0.139] [39.894] [0.114]
Beta 952 3SLS Robust Error
0.132*** -0.037 2.548*** 0.044 3.219*** 0.267*
[0.035] [0.030] [0.625] [0.120] [0.917] [0.150]
14 3SLS Fama-McBeth 0.061** 0.013 1.182** 0.127 1.161* 0.369**
[0.025] [0.030] [0.469] [0.096] [0.574] [0.132]
Idiosyn_Risk
952 3SLS Robust Error
0.000 0.000 97.008*** 0.005 97.992*** 0.609***
[0.000] [0.000] [38.413] [0.118] [31.691] [0.081]
14
3SLS Fama-McBeth -0.0003 0.001 11.053 0.133 86.981* 0.511***
[0.001] [0.001] [72.632] [0.153] [40.996] [0.103]
All variable definitions are in Appendix A.
41
Figure 1
Mean CEO Compensation: Banks and Non-Banks
$0
$2,500
$5,000
$7,500
$10,000
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
$ t
ho
usan
ds
banks
non-banks
Figure 2
Mean Vega: Banks and Non-Banks
0
100
200
300
400
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
veg
a banks
non-banks
42
Figure 3
Mean Delta: Banks and Non-Banks
0
500
1000
1500
20001992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
delt
a banks
non-banks
Figure 4
43
Appendix A
Variable Definitions
Panel A: CEO Characteristics
Vega The pay-risk sensitivity, which is the change in the dollar value of CEO wealth for a
0.01 change in stock return volatility, measured by partial derivatives of Black-Scholes
value of options with respect to stock return volatility.
Delta The pay-performance sensitivity, which is the change in the dollar value of CEO wealth
for a 1% change in stock price, measured by partial derivatives of Black-Scholes value
of options and market value of stock holdings with respect to stock price.
Salary The CEO’s annual base salary.
Bonus The CEO’s annual bonus.
Option_Grants The CEO’s total value of stock options granted during the year (using Black-Scholes).
Rstock_Grants The CEO’s total value of restricted stock granted during the year.
Total_Compensation The CEO’s total annual compensation comprised of the following: salary, bonus, total
value of stock options granted, total value of restricted stock granted, long-term
incentive payouts, and all other total. Tenure The number of years in the CEO’s term.
Panel B: Business Policy Measures
Nonint Total noninterest income, scaled by net operating income.
Nonint_Less Total noninterest income less fiduciary income and deposit service charges, scaled by
net operating income.
Nonint_Sec Noninterest income from loan securitization and servicing, scaled by net operating
income.
Nonint_Nontrad Noninterest income from nontraditional banking activities such as trading revenue,
venture capital revenue, and underwriting income from insurance activities, scaled by
net operating income.
Loans Total loans and leases held in portfolio, scaled by total assets.
Commer_Loans Commercial and industrial loans, scaled by total assets.
Commer_Real_Loans Commercial real estate loans, scaled by total assets.
Alloc_loans Allocation for loan and lease losses, scaled by total assets.
Prov_loans Provision for loan and lease losses, scaled by total assets.
Charge_loans Charge-offs and recoveries on loan and lease losses, scaled by total assets.
Noncore_Deposit Noncore deposits such as time deposits of $100,000 or more, scaled by total assets.
Short_Funds Federal funds purchased and securities sold under agreements to repurchase, scaled by
total assets.
Equity_Mult Total assets over total equity capital.
Equity_Mult_Flows The sum of total assets and off-balance sheet items such as unused commitments and
letters of credit over total equity capital.
Equity_Mult_Off Net operating income over total equity capital.
Growth The natural log of ending total assets over beginning total assets.
Intangible Intangible assets, scaled by total assets.
Deriv_Notional The notional amount of derivative contracts held for trading purposes, scaled by total
assets.
Deriv_FV The fair value of derivative contracts held for trading purposes, scaled by total assets.
MBS_Private_BV The amortized cost of private mortgage backed securities, scaled by total assets.
MBS_Private_FV The fair value of private mortgage backed securities, scaled by total assets.
MBS_Struct_BV The amortized cost of structured mortgage backed securities, scaled by total assets.
MBS_Struct_FV The fair value of structured mortgage backed securities, scaled by total assets.
MBS_Pass_BV The amortized cost of pass-through mortgage backed securities, scaled by total assets.
MBS_Pass_FV The fair value of pass-through mortgage backed securities, scaled by total assets.
Risk The standard deviation of daily stock returns over a year.
Beta The beta coefficient estimated from the market model over a year.
Idiosyn_Risk The standard deviation of the market model residuals over a year.
44
Appendix (Continued)
Panel C: Bank Characteristics
Net_Oper_Rev The sum of net interest income and total noninterest income.
Total_Assets The ending balance of total assets.
ln(TA) The natural log of beginning total assets.
MB The beginning balance of the market-to-book ratio of equity.
ln(MB) The natural log of the beginning market-to-book ratio of equity.
Cash_Balance Cash and balances due from depository institutions, scaled by total assets.
Deposit Total deposits, scaled by total assets.
Core_deposit Total deposits less noncore deposits, scaled by total assets.
Equity_Ratio Total equity capital over total assets.
Net_Interest_Margin The difference between interest and fee income on loans over total loans and interest on
deposits over total deposits.
Foreign_Deposit Deposits in foreign offices, edge and agreement subsidiaries and IBFs, scaled by total
assets.
One_Yr_Gap The level of repricing asymmetry between assets and liabilities, scaled by total assets.
Nonint_Less/Beg(TA) Noninterest income less fiduciary income and deposit service charges, scaled by
beginning total assets.
Trading_Rev/Beg(TA) Trading revenue over beginning total assets.
45
Appendix B: Identifiers in the Business Policy Regressions
Policy Measures Identifiers in the policy regression
Nonint Short_Fundst-1, Depositt-1, Loanst-1
Nonint_Less Short_Fundst-1, Depositt-1, Loanst-1
Nonint_Sec Short_Fundst-1, Depositt-1, Loanst-1
Nonint_Nontrad Short_Fundst-1, Depositt-1, Loanst-1
Loans Core_Depositt-1, Equity_Ratiot-1
Commer_Loans Core_Depositt-1, Equity_Ratiot-1
Commer_Real_Loans Core_Depositt-1, Equity_Ratiot-1
Alloc_loans Loanst
Prov_loans Loanst
Charge_loans Loanst
Noncore_Deposit Nonint_Less/Beg(TA)t, Growtht-1
Short_Funds Nonint_Less/Beg(TA)t, Growtht-1
Equity_Mult Nonint_Less/Beg(TA)t
Equity_Mult_Flows Nonint_Less/Beg(TA)t
Equity_Mult_Off Nonint_Less/Beg(TA)t
Growth Core_Depositt-1, Equity_Ratiot-1 , Net_Interest_Margint-1
Intangible Core_Depositt-1, Growtht-1
Deriv_Notional Foreign_Depositt-1, One_Yr_Gapt-1 , Trading_Rev/Beg(TA)t
Deriv_FV Foreign_Depositt-1, One_Yr_Gapt-1 , Trading_Rev/Beg(TA)t
MBS_Private_BV Short_Fundst-1, Depositt-1
MBS_Private_FV Short_Fundst-1, Depositt-1
MBS_Struct_BV Short_Fundst-1, Depositt-1
MBS_Struct_FV Short_Fundst-1, Depositt-1
MBS_Pass_BV Short_Fundst-1, Depositt-1, Loanst-1
MBS_Pass_FV Short_Fundst-1, Depositt-1, Loanst-1
Risk Loanst-1, Nonint_Less/Beg(TA)t, Equity_Ratiot-1
Beta Loanst-1, Nonint_Less/Beg(TA)t, Equity_Ratiot-1
Idiosyn_Risk Loanst-1, Nonint_Less/Beg(TA)t, Equity_Ratiot-1
46
Appendix C: Full Regression Results for Nonint
Nonintt Nonintt Nonintt Ln(Vegat) Ln(Deltat) Nonintt Ln(Vegat) Ln(Deltat)
OLS robust
Error
OLS Fama
McBeth 3SLS Robust Error 3SLS Fama McBeth
ln(Vegat-1) 0.023*** 0.0217***
[0.004] [0.002]
ln(Deltat-1) -0.011** -0.006
[0.005] [0.006]
ln(Vegat) 0.102*** 0.610*** 0.104*** 0.556***
[0.018] [0.070] [0.022] [0.117]
ln(Deltat) -0.098*** 0.064 -0.077*** 0.155
[0.016] [0.124] [0.019] [0.145]
Nonint_Incomet 1.241*** 1.077** 1.579*** 1.461
[0.475] [0.496] [0.544] [0.853]
ln(TA t-1) 0.022*** 0.0204*** 0.011 0.284*** 0.067 0.002 0.208*** 0.076
[0.005] [0.005] [0.010] [0.059] [0.052] [0.009] [0.068] [0.060]
ln(MBt-1) 0.235*** 0.2606*** 0.301*** -0.036*** 1.037*** 0.332*** -0.281 1.137***
[0.023] [0.024] [0.030] [0.285] [0.175] [0.027] [0.358] [0.263]
Short_Fundst-1 -0.408*** -0.317** -0.364*** -0.196
[0.124] [0.116] [0.098] [0.128]
Depositt-1 -0.542*** -0.538*** -0.535*** -0.522***
[0.065] [0.066] [0.065] [0.071]
Loant-1 -0.387*** -0.397*** -0.349*** -0.336***
[0.037] [0.028] [0.036] [0.043]
Ln(Salaryt-1) 1.370*** 1.101***
[0.206] [0.211]
Tenuret 0.052*** 0.050***
[0.005] [0.012]
Cash_Balancet-1 -4.204*** -3.388***
[0.771] [0.676]
Year Dummies Yes Yes Yes Yes
Annual Regressions 14 14 14 14
Observations 773 773 926 926 926 926 926 926